August 2016 - Darwin-kernel

Re: Kext symbolizing & missing kernel slide
by Quinn "The Eskimo!" 29 Aug '16

29 Aug '16

On 26 Aug 2016, at 16:59, Andrew Keller <andrew(a)kellerfarm.com> wrote: > Should we assume that a static const variable locally declared in a function is never stored on the stack? Yes, that’s a reasonable assumption. Share and Enjoy -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Relations, Developer Technical Support, Core OS/Hardware _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Unit testing kexts
by Stan Sieler 26 Aug '16

26 Aug '16

Re: > space application and performing our tests there. Over time, as we have increased our code coverage of the unit tests, we’re noticing a pretty strong upward trend of the number of compiler conditionals required to support cross-compiling between a kext and a user-space app. For example: ... #ifdef UNIT_TEST > str = (char*) malloc(sz); > #else > str = (char*) OSMalloc(sz, gOSMallocTag); > #endif ... > Suffice it to say, we’re wondering if there’s a better way. Ideally, we would like to unit test the code that’s actually running in production. These compiler-level conditionals force us to essentially maintain two versions of our code. Worse than that, the unit tested version is not the production version. The following is not a kext/Mac/platform-specific suggestion, but... When I encounter this, I move the code into a (hopefully small) module of routines that tend to vary (or not exist) from platform to platform and then use "cover functions" to access them. E.g., str = ss_malloc (sz); ...which gets implemented as one of the two different malloc's you show. So, my "ss_read" and "ss_seek", for example, take a guaranteed int64_t #bytes, and returns a guaranteed int64_t result ... no worrying about int vs. size_t vs. ssize_t vs. offset_t, for example. And, no worrying about whether I need to call read or read64 (or lseek / lseek64), which I used to have to worry about on some platforms. (Inspired, years ago, by Terry Winograd's "Breaking The Complexity Barrier, Again" article :) Stan _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Kext symbolizing & missing kernel slide
by Andrew Keller 26 Aug '16

26 Aug '16

_______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Kext symbolizing & missing kernel slide
by Quinn "The Eskimo!" 26 Aug '16

26 Aug '16

On 26 Aug 2016, at 03:21, Andrew Keller <andrew(a)kellerfarm.com> wrote: > Perhaps that’s the machine code that pushes the function's local variables onto the stack? Most likely. > Is all memory required for all local variables in a given function allocated on the stack immediately upon invoking the function … ? Yes. Well, it’s not /guaranteed/ to be that way—in most situations the exact mechanics of this is up to the compiler—but the general approach taken by most compilers is to adjust the stacker pointer once in the function prologue [1] and leave it there until the function returns. > A macro that allocates a 4KB array on the stack every time it’s used. Yowsers! 4 KiB on a user stack is a lot; it’s crazy talk on a kernel stack (-: > Fortunately, that macro represents a goldmine of opportunities for improvement. Yep. Share and Enjoy -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Relations, Developer Technical Support, Core OS/Hardware [1] <https://en.wikipedia.org/wiki/Function_prologue> _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Kext symbolizing & missing kernel slide
by Andrew Keller 26 Aug '16

26 Aug '16

Am 25.08.2016 um 6:22 nachm. schrieb Brian Bechtel <brian.bechtel(a)gmail.com>: > Apple implemented kernel address slide randomization in 10.8. Assume a slide of zero for 10.7 and earlier. Wow! Didn’t know that. With that, I was able to pretty confidently symbolize the backtrace, and it points to one of the lines of assembly before the first line of my code in a function. Perhaps that’s the machine code that pushes the function's local variables onto the stack? If all stack memory for a function is allocated immediately (question about that below), then it would not be surprising that the stack overflow happens when the function loads all of its local variables onto the stack, before the first line of my code. > Double fault panics usually indicate you are using too much of the very limited 16K kernel stack space. Look for allocation of large arrays on the stack. Is all memory required for all local variables in a given function allocated on the stack immediately upon invoking the function, or are local variables dynamically created and destroyed on the stack as the function drags on and variables transition in and out of scope? If you say it’s the former, then I have a suspect — A macro that allocates a 4KB array on the stack every time it’s used. In the entire call stack leading up to the panic, there are 3 usages of that macro. All it would take is 4KB of other stuff on the stack for something to go _boom_. Fortunately, that macro represents a goldmine of opportunities for improvement. It’s straight-forward, relatively isolated, and easy to change without risking breaking other things. Thanks, - Andrew Keller _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Kext symbolizing & missing kernel slide
by Brian Bechtel 25 Aug '16

25 Aug '16

_______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Kext symbolizing & missing kernel slide
by Andrew Keller 25 Aug '16

25 Aug '16

_______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Unit testing kexts
by Thomas Tempelmann 25 Aug '16

25 Aug '16

_______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Unit testing kexts
by Andrew Keller 25 Aug '16

25 Aug '16

_______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Here is the detail of configuring PanicDump
by Prokash Sinha 24 Aug '16

24 Aug '16

Configure the server side 738 ls /PanicDumps/ 739 sudo chown root:wheel /PanicDumps 740 sudo chmod 1777 /PanicDumps 741 sudo launchctl load -w /System/Library/LaunchDaemons/com.apple.kdumpd.plist 742 sudo launchctl list | grep dump — to get that kdumpd is enabled 743 ifconfig -- to get the ip to be used in the client — Configure the client CNTL+R or CMD+R to get to reboot on recovery mode. once in recovery mode, open terminal from utility menu. # csrutil disable #reboot --- Now configure the client # sudo nvram boot-args="debug=0xd44 _panicd_ip=<ip of the server>” # reboot # nvram boot-args Or even # sudo nvram boot-args="debug=0xd44 _panicd_ip=<ip of the server> kdp_match_name=en0 _router_ip=<router IP addr>” reboot $sudo dtrace -w -n “BEGIN{ panic();}” Takes the client to Halt mode, but don’t see any progress bar of data being transferred. Wireshark/pcap does not show anything. Server to client - PING 10.34.200.171 (10.34.200.171): 56 data bytes 64 bytes from 10.34.200.171: icmp_seq=0 ttl=64 time=165.644 ms 64 bytes from 10.34.200.171: icmp_seq=1 ttl=64 time=820.634 ms 64 bytes from 10.34.200.171: icmp_seq=2 ttl=64 time=8.032 ms 64 bytes from 10.34.200.171: icmp_seq=3 ttl=64 time=569.325 ms 64 bytes from 10.34.200.171: icmp_seq=4 ttl=64 time=1.223 ms 64 bytes from 10.34.200.171: icmp_seq=5 ttl=64 time=91.462 ms 64 bytes from 10.34.200.171: icmp_seq=6 ttl=64 time=1.321 ms 64 bytes from 10.34.200.171: icmp_seq=7 ttl=64 time=213.236 ms 64 bytes from 10.34.200.171: icmp_seq=8 ttl=64 time=27.520 ms Request timeout for icmp_seq 9 64 bytes from 10.34.200.171: icmp_seq=9 ttl=64 time=1127.781 ms Likewise client to server ping goes fine. I just want to see if there is a way to find out if the packets are being sent to server for the dump data. What, if any I am missing. TIA, Prokash _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0